Incomplete Fusion in Welding: Causes, Detection, Prevention & Repair

Table of Contents

What Is Incomplete Fusion in Welding? A Complete Introduction

The weld looked perfect. Uniform ripples, no porosity, no undercut — it passed every visual check on the shop floor. Eighteen months later, the pressure vessel it held together was leaking along a hair-thin unbonded seam that had been sitting inside the joint the entire time.

That invisible defect is incomplete fusion, and it is one of the most common reasons welds fail in service. It does not announce itself. It hides under a bead that looks flawless, and it waits.

What Incomplete Fusion Actually Means

Incomplete fusion — also called lack of fusion — is a weld discontinuity in which the weld metal does not fuse to the base metal, to the joint sidewall, or to a previously deposited bead. The filler metal melted. The arc ran. A bead formed. But at the interface, no true metallurgical bond ever developed.

Standards bodies treat the term precisely. AWS A3.0M/A3.0 Standard Terms and Definitions defines the vocabulary inspectors and welders use when classifying weld discontinuities such as incomplete fusion. Internationally, ISO 6520-1 sorts lack of fusion into defect group 401, with subtypes for sidewall (4011), bead-to-bead or interpass (4012), and root (4013). Those numbers matter on an inspection report, because each location points back to a different root cause.

Where Incomplete Fusion Hides Inside a Joint

Fusion can fail at four different interfaces, and each one behaves differently in service:

Sidewall fusion — the weld metal never bonds to the bevel face or the leg of a fillet. Thick sections and narrow grooves make this the most frequent form.
Interpass (bead-to-bead) fusion — a new pass lays over a previous bead without bonding to it, usually after poor cleaning, low heat input, or a wandering arc.
Root fusion — the first pass fails to bond at the root face, leaving a sharp unfused line at the very bottom of the joint.
Weld toe fusion — the edge where the bead meets the base plate stays unbonded, often just below the surface where no eye can reach.

Fusion vs. Adhesion: Why It Is Planar and Sharp

True fusion is a metallurgical event, not a contact sport. The base metal and filler must both melt at the interface, mix through dilution, and solidify as one continuous structure. When that happens, grains grow across the boundary and the joint becomes a single piece of steel. When it does not, the weld metal is merely sitting on top of the base plate — adhered, not bonded.

Adhesion is the cruelest kind of defect because it looks exactly like fusion from the outside. Because the unbonded interface has essentially zero thickness, incomplete fusion is planar — a flat, sharp-edged crack waiting to open. That geometry is why it concentrates stress far more severely than rounded discontinuities such as porosity or slag inclusions.

Why a Sharp Unbonded Line Becomes a Crack

A planar defect behaves like a pre-made crack. Under cyclic loading, the sharp tip raises local stress enormously — far above the nominal stress the design engineer calculated — so fatigue cracks initiate there long before the rest of the weld is stressed. In pressure vessels, tanks, and structural connections, that unbonded seam is the starting line for failure.

Picture a welder running a 3/8-inch fillet on half-inch plate with a wire welder. The machine is set a little low, travel speed creeps up, and the arc rides the top of the puddle instead of digging into the corner. The bead that comes out is smooth and even. The sidewall underneath never melted. Nothing about that weld looks wrong — which is exactly the problem, and the reason so many good-looking welds fail inspection when they are cut and etched.

Quick Note: Incomplete fusion is one of the few defects you cannot reliably judge by eye. Most welding codes, including AWS D1.1, treat lack of fusion as rejectable in the vast majority of applications — no size allowance, no gray area.

Whether you are picking up a stinger for the first time, inspecting welds to code, or managing a fabrication shop, this guide is for you. We will cover what causes incomplete fusion, how to detect it with visual and NDT methods, how to prevent it process by process, and how to repair it correctly.

Incomplete Fusion vs. Lack of Penetration vs. Cold Lap: Untangling the Terminology

Say “incomplete fusion” in one shop and you’ll get a knowing nod. Say it in another, and someone will insist you actually mean lack of penetration. The confusion is genuine, and it costs money: an inspector writes up the wrong defect, a welder chases the wrong variable, and a repair gets ground out of the wrong location.

Three different problems hide behind these labels. They overlap in search results, in shop talk, and in the code book — but they are not the same defect, and they do not share the same fix.

Lack of Fusion (Incomplete Fusion)

This is a bonding failure. Weld metal sits against base metal or a previous bead, yet no metallurgical bond formed across that interface. The two surfaces touched and cooled without ever melting together. “Lack of fusion” and “incomplete fusion” are the same defect in most U.S. shop talk, and AWS D1.1 treats it as a planar, crack-like discontinuity — which is exactly why it gets rejected so fast. Typical locations: sidewalls, the weld toe, between passes, and the root of a partial-joint detail.

Incomplete Joint Penetration (IP)

This one is geometric, not a bonding issue. The weld is soundly bonded, but it does not extend through the full required joint thickness. What is left unbonded is unmelted base metal at the root — a root land the arc never reached. Root gap, root face, bevel angle, and travel speed drive it. A sound weld can still be an incomplete-penetration weld.

Cold Lap (Cold Shut)

Cold lap is the colloquial name for a fusion-type defect where molten weld metal rolls over an unmelted ledge at the toe or sidewall, draping across base metal without bonding. It appears most often in short-circuit GMAW. Codes classify it by morphology — usually as incomplete fusion, occasionally as overlap, depending on how the bead sits. To tell true overlap apart from a fusion defect, our guide to the overlap welding defect walks through the geometry.

	Lack of Fusion	Incomplete Penetration	Cold Lap
Unbonded feature	Weld-to-base or weld-to-bead interface	Root land (unmelted base metal)	Weld toe or sidewall ledge
Typical location	Sidewalls, toe, between passes	Root of partial-joint welds	Toe, sidewall
Root cause	Low heat input, poor angle, fast travel	Poor fit-up, small gap, large root face	Cold puddle, low voltage, wrong angle
Best NDT	UT; RT is limited	RT, root-side visual	Visual, UT
Code treatment	Planar defect; reject	Geometric shortfall; reject	Classified by morphology

Radiography, meanwhile, is notoriously weak against tight planar lack-of-fusion. The X-ray beam passes through the unbonded interface with almost no change in density, so a real defect can leave nothing on the film. Ultrasonic testing is far more reliable here. Incomplete penetration, being a genuine geometric gap, is much easier for RT to flag.

Miller Electric notes that lack of fusion in MIG welding comes primarily from improper gun angle or incorrect travel speed, and from insufficient heat in the weld. Notice what that list does not include: root gap. Fusion is a heat-and-technique problem, and misreading it as a penetration problem sends a welder adjusting fit-up while the real culprit — low heat input — stays untouched. Building your eye for the difference takes reps; good and bad MIG welds is a good place to start.

One more wrinkle: what an inspector calls the defect often depends on morphology, not intent. A toe defect that overlaps base metal but shows a fusion line gets written up one way; the identical geometry with a sharp unfused interface gets written up another. The practical takeaway is the same either way — the bond never formed.

Quick Note: Root-of-joint defect in a properly fitted joint? Think penetration. Sidewall or interpass defect with correct fit-up? Think fusion. The location of the unbonded interface tells you which variable to reach for.

The bench rule: before you touch a single dial, find where the defect lives. Root land means penetration. Sidewall, toe, or interpass means fusion. Treat any suspected fusion defect as crack-like until UT says otherwise. Wrong diagnosis, wrong fix.

What Causes Incomplete Fusion? Root Causes by Welding Process

Ask five welders why a joint refused to fuse and you’ll hear five answers — a dirty plate, a cold machine, “the wire was acting up.” Underneath the shop-floor theories, almost every case traces back to the same physics: the arc didn’t deliver enough energy to melt both the filler and the base metal at the interface, or something physically blocked that melting. The drivers cluster into four families, and the process you’re running decides which one bites hardest.

The Four Families of Root Causes

Too little heat. Amperage or voltage set too low, wire feed speed mismatched to voltage, travel speed too fast, wrong polarity, or an electrode too large for the joint. Insufficient heat input relative to joint thickness is the single most common driver of sidewall and interpass lack of fusion.
Poor joint access and geometry. A narrow groove, a thick root face, a tight root gap, or a deep sidewall the arc simply cannot reach and wet. Electrode angle makes it worse — point the arc at the wrong wall and the opposite side stays cold.
Contamination and surface condition. Mill scale, rust, oil, paint, moisture, and hydrated aluminum oxide all sit between the arc and the metal, preventing the base metal from melting and wetting at the joint interface. Insufficient interpass cleaning and slag left on the joint do the same.
Technique and process characteristics. Long arc length, weaving that is too wide or too fast, short-circuit transfer run cold, insufficient preheat on thick or high-carbon and alloy steel, the wrong shielding-gas mix, and a worn contact tip or bad liner that disrupts current transfer.

Quick Note: Heat input is easy to estimate: kJ/in = (volts × amps × 60) ÷ (1000 × travel speed in inches per minute). Slowing travel raises heat input and helps wet the sidewalls — but dwell too long, especially in GMAW short-circuit transfer, and the puddle piles up and rides over cold base metal, producing cold lap instead of fusion.

These families rarely act alone. A slightly undersized electrode, a joint that’s too tight to reach the root, and a plate that still has mill scale can stack into a weld that looks great and fuses nowhere. That’s why troubleshooting starts with the process — each one fails in a characteristic way.

Where Each Process Goes Wrong

GMAW. Cold lap at the weld toes is the signature failure, usually from voltage set too low or travel too fast. A worn or wrong contact tip also wrecks current transfer and stubs the arc out of position. Learners chasing a pretty bead often raise travel speed until the toes stop wetting; our comparison of good and bad MIG welds shows what those cold toes actually look like.

FCAW. Slag left between passes traps the next bead on top of a contaminant layer. Left in place, that slag becomes the bond line the next pass never crosses. Self-shielded flux-cored wires require direct current electrode negative (straight polarity); running them on the wrong polarity changes the arc characteristics and fusion behavior — see our guide to DCEN vs DCEP.

SMAW. Fast travel, low current, a bad rod angle, and arc blow are the usual suspects. E7018 demands enough amperage for the rod diameter; undershoot it and the joint never gets above the fusion line. Our 3/16 E7018 amperage settings article breaks down the right ranges.

TIG. Insufficient heat, no filler dabbed into the sidewall, and oxidized aluminum all prevent wetting. Polarity matters enormously; our guide to TIG welding polarity breaks down how AC balance and DCEN/DCEP trade cleaning against penetration.

SAW. Flux/bead misalignment and insufficient current on heavy plate bury the arc without fusing the root or sidewall. On thick sections, the current that’s fine for thin plate is nowhere near enough — the robotic SAW setups used in heavy fabrication run far higher amperage for exactly this reason.

Robotic and automated welding. Programming errors, tool center point drift, and joint variability all push the arc off the intended line. Joint variability is the sneaky one: as fit-up drifts, the weld program keeps running the same path into a gap that moved.

Note: Don’t chase fusion by burying your face in a cold, smoky arc. OSHA flags welding fume exposures — including manganese — as a genuine health hazard, so fix the settings rather than the distance.

For a broader defect reference, Miller’s weld troubleshooting guide groups lack of fusion alongside its related discontinuities.

Whichever process you run, the diagnosis is the same: heat, access, cleanliness, and technique — in that order.

How to Detect Incomplete Fusion: Visual, NDT, and Destructive Methods

Incomplete fusion is a liar of a defect. It hides inside the joint, often as a tight, planar unbonded interface only microns wide, with no gap, no porosity, and no crack to open up. That geometry is exactly why it slips past so many inspections — and why the method you choose matters more than how many passes you run. Detection isn’t a single test; it’s a planned mix matched to joint geometry, material, and criticality.

Visual Inspection (VT): Good for Symptoms, Blind to Sidewalls

VT is the fastest, cheapest check on the floor and always the first. It’s excellent at catching cold lap at the weld toe, ragged or wandering toes, an excessively convex bead, uneven leg lengths, and the kind of too-shallow penetration profile that hints the arc never reached the root. Inspectors look at toe lines, ripple uniformity, spatter patterns, and whether the cap sits high and narrow instead of blending smoothly into the base metal.

What VT cannot do is see sidewall fusion. A beautiful, uniform cap tells you the welder had steady hands — nothing about whether the arc melted the joint faces. Remember, good-looking welds and bad MIG welds can look deceptively similar, and the difference often lives below the surface.

Liquid Penetrant (PT) and Magnetic Particle (MT)

PT only works when the defect breaks the surface and can hold dye. A tight internal LOF won’t hold anything, so PT often gives a false clean. MT is the stronger of the two for fusion defects: in ferromagnetic base metals it reveals surface and near-surface planar discontinuities, and lack of fusion is planar. It’s a solid screening tool on carbon steel, but it fades out as the flaw gets deeper.

Radiography (RT): The Default That Disappoints

RT has long been the default for code welds because it produces a permanent record. But tight planar defects are its weakness: a bond line with no gap and no density gradient across it simply doesn’t show the contrast change RT needs. That’s the key fact to internalize — lack of fusion with tight, planar geometry is generally more reliably detected by ultrasonic testing than by radiography. Radiographers can improve results with angle-shot techniques, but no amount of film interpretation fixes a defect with nothing to image.

Short Note: RT can pass a weld that UT later flags, and that a service failure eventually confirms. Never treat a clean radiograph as proof of full fusion on a planar-critical joint.

Ultrasonic Testing (UT), Phased Array, and TOFD: The Planar-Defect Specialists

UT works by bouncing sound off an interface, so it detects the flaw itself rather than a density difference. Angle-beam shear-wave scanning is the classic approach, and phased-array UT (PAUT) plus time-of-flight diffraction (TOFD) have made it faster and better at sizing. These methods shine on internal planar discontinuities oriented roughly perpendicular to the plate surface — exactly how sidewall and root LOF present. That’s why UT is now standard practice on many structural and pipeline jobs, and why automated lines increasingly pair welding with in-process monitoring. For high-deposition processes like robotic submerged arc welding, UT is the practical backstop. Limits: UT needs skilled operators, good surface condition, and careful calibration, and it struggles on coarse-grained austenitic stainless without special techniques.

Destructive Testing: Where LOF Gets Caught Most Often

Here’s the honest answer to “where does incomplete fusion usually get found?” — on test coupons, not production welds. Guided bend and macro-etch tests on procedure qualification coupons are standard ways to reveal lack of fusion during qualification under AWS D1.1 and ASME BPVC Section IX practice. A bend coupon pulls the joint faces apart, macro etch shows the bond line in cross-section, nick-break tests open the root, and tensile plus metallographic examination confirm fusion at depth. This is why qualification coupons and welder performance tests remain the best place to catch lack of fusion before a part ships.

Quick Note: Building real inspection judgment takes practice — the same way welding practice builds torch control. AWS Certified Welding Inspectors (CWIs) are trained in discontinuities, inspection, and acceptance criteria, and their job is to tie every method to the applicable code and the joint’s service demand.

Why Incomplete Fusion Is So Dangerous: Code Rules and Real-World Consequences

Incomplete fusion rarely announces itself. Unlike porosity — rounded, blunt voids that redistribute stress relatively gently — lack of fusion is a planar discontinuity. Its unbonded tip is sharp, sometimes nearly atomically tight, so the stress field at that tip piles up dramatically. In fracture-mechanics terms, lack of fusion behaves much like a crack: an efficient stress concentrator waiting for the right mix of load cycles, restraint, and low temperature to begin growing. That is the core reason inspectors and engineers treat it as a crack-like flaw rather than a cosmetic blemish on a weld.

From Discontinuity to Crack: Why the Tip Matters

Under static tension, a planar lack-of-fusion defect can sit quietly for years. Add cyclic loading — traffic vibration on a bridge, pressure swings in a vessel, the repeated lift cycles of a crane boom — and that sharp tip becomes a fatigue initiation site long before the surrounding sound weld would ever crack. Drop the temperature and the picture worsens: as base-metal notch toughness falls, a defect that might tear slowly in summer can drive a brittle fracture through an entire restrained weldment in milliseconds. This is why AWS D1.1/D1.1M, the structural welding code for steel, and the fracture-control philosophies behind pipeline standards refuse to treat fusion defects casually.

Real-World Consequences: Loss of Containment, Rework, and Liability

The consequences ripple outward from the joint. In pressure equipment, a lack-of-fusion seam is a ready-made leak path; in pipelines and tanks, it means loss of containment, environmental release, and shutdown. Beyond the human cost, the economics are brutal: a rejected fabrication must be cut out, re-welded, and re-inspected, and every repair introduces fresh heat input, residual stress, and distortion into metal that was already acceptable. Add schedule slippage, third-party re-inspection, and potential liability or personal injury, and one hidden unbonded interface can cost far more than the entire original weld.

Pressure vessels and tanks: unbonded seams become leak paths under thermal and pressure cycling.
Bridges and crane booms: cyclic loading turns the sharp tip into a fatigue crack origin.
Pipelines: containment loss with environmental and regulatory fallout.

High-deposition processes are not exempt — even robotic submerged arc welding on thick plate depends on genuine sidewall fusion to be sound.

Acceptable vs. Rejectable: The Spec Always Answers the Question

This is the part welders and inspectors argue about most. Fusion-related acceptance limits are code- and service-specific. AWS D1.1/D1.1M defines what is rejectable in structural steel; ASME BPVC Section IX governs welding procedure and performance qualification for pressure equipment, while B31.3 and Section VIII supply the fabrication acceptance criteria. API 1104 is the widely used standard for pipeline welding, with its own radiographic acceptance criteria for fusion-type defects. The same indication can be perfectly acceptable in one application and rejectable in another, so “is this OK?” must always be answered against the governing specification — never by opinion.

A Failure Scenario: It Passed Visual and Radiography

Consider a fabricated crane boom section. Visual inspection found uniform beads, no undercut, no porosity. Radiography looked clean. Eighteen months of service later, a crack ran from a sidewall lack-of-fusion defect straight through the boom. The cause was simple: RT does not always resolve a tight, unbonded planar interface, because the beam passes through the thin gap without enough density change to register. Planar defects are inspection-method dependent — lack of fusion and overlap demand techniques chosen deliberately for their orientation, often ultrasonic or phased-array testing rather than film alone.

Quick Note: A defect that passes radiography can still be a serious lack-of-fusion discontinuity. RT is weakest against tight, planar, unfused interfaces that sit close to the beam direction — the flaw that looks clean on film may be the one that cracks in fatigue.

The Regulatory and Safety Frame

OSHA requires employers to address welding, cutting, and brazing hazards across general industry, maritime, and construction standards, and its hazard guidance covers fumes, UV radiation, and burns, while NIOSH adds detailed material on fume exposure. A fusion defect is a durability problem, but the arc itself is an immediate safety problem — a safe shop treats both with the same discipline.

Fixing Incomplete Fusion Process by Process: GMAW, FCAW, SMAW, TIG, and SAW

Fusion defects obey the same physics in every process: enough heat, in the right place, for long enough. What changes is which knobs you turn. Here is the correction path for each process, built around heat input, filler choice, polarity, travel, and joint prep.

GMAW: Add Heat, Shorten Stickout, Aim at the Leading Edge

In MIG, lack of fusion almost always means the puddle is cold or the arc is pointed at the wrong spot. Work through it in this order:

Heat input: raise voltage or wire feed speed to add heat and eliminate the cold-weld condition that starves sidewall fusion. With DCEP, about two-thirds of the arc heat goes into the workpiece; flip to DCEN and that balance shifts toward the electrode.
Angle and aim: Miller Electric recommends a 0- to 15-degree gun angle and keeping the arc on the leading edge of the weld pool — aiming at the middle or trailing edge just drapes filler over cold steel.
Stickout: shorten it. Excess stickout drops voltage and amperage at the arc, and the weld goes cold.
Transfer and consumables: use spray or pulsed transfer on thicker sections, and check that your contact tip is clean and unworn so current transfer stays stable.

Finally, scrub mill scale, rust, and paint from the fusion faces. No arc can fuse to contamination.

Quick Note: Change one variable at a time. Raise heat first, then fix aim, then travel speed. Chasing three settings at once turns a simple fix into guesswork.

FCAW: Confirm Polarity, Then Kill the Slag

Flux-cored fusion failures are usually polarity mistakes or slag problems. Verify polarity first: gas-shielded wires typically run DCEP, while self-shielded wires run DCEN. Reversing this changes penetration and wetting dramatically, so review the polarity for flux core welding guidance and the wire data sheet. Then:

Remove slag completely between passes — trapped slag becomes an unbonded line.
Watch for “wagon tracks,” the parallel grooves that signal slag or unfused sidewall. Stop and repair immediately.
Adjust travel speed and work angle so the pool wets both toes, and bump voltage slightly if the puddle is sluggish.

SMAW: Amperage, Arc Length, and Dry Rods

Stick welding rewards force and precision. Increase amperage within the electrode maker’s range, shorten arc length, and keep the arc biting into the root and both sidewalls instead of racing ahead. Slow down when the puddle refuses to wet out.

Arc blow pushes the arc away from the joint and leaves one sidewall cold — shorten the arc, reduce current, reposition the work lead, or use a whip technique. Electrode moisture degrades arc behavior too: E7018 is low-hydrogen and moisture-sensitive, so keep rods in a heated oven and discard exposed ones. Size-specific starting points live in guides like 3/16 E7018 amperage.

TIG: Raise Current and Wet Both Sidewalls

TIG fusion is a heat-balance act. Raise current until the pool wets both sidewalls and looks fluid, then add filler at the leading edge so it fuses into the root instead of sitting on top. On aluminum, the surface oxide melts near 3,700°F (about 2,055°C) while the metal beneath melts around 1,220°F (about 660°C) — that gap is why cathodic cleaning matters. Bias the AC balance toward more cleaning and control interpass temperature on thick sections. The full polarity picture is in TIG welding polarity.

SAW: Procedure Discipline Above All

Submerged arc fusion depends on wire–flux compatibility and steady current density. Confirm the flux hopper flows evenly, the wire grade matches the procedure, the bead tracks the joint centerline, and travel speed holds constant. On heavy plate, verify sidewall penetration with a qualified procedure — automated work such as robotic submerged arc welding depends entirely on qualified parameters. Travel too fast, or an arc riding one side of the joint, produces exactly the unbonded walls this article is about.

Troubleshooting Quick Reference

Symptom	Likely cause	Fix
Toe-line cold lap	Arc aimed at puddle center, travel too fast, low heat	Raise voltage or WFS; hold 0–15° angle on the leading edge
Sidewall gaps in a groove	Poor fit-up, low current, torch pulled off the wall	Re-fit the joint, raise current, pause briefly at each sidewall
Interpass unbonded line	Slag not removed, cold pass over cold metal	Chip and brush between passes; raise interpass heat input

Practical Tips: Joint Prep, Fit-Up, Travel Speed, and Working Position

Most incomplete fusion is created before the arc ever strikes. By the time you pull the trigger, the outcome is largely decided — by the gap you left, the scale you didn’t grind, the angle you can’t reach. Fix the front end and the weld takes care of itself.

Fit-Up: The Root Gap Does the Heavy Lifting

A root gap that is too tight, or a root face that is too thick, physically blocks the arc from reaching the root. The arc can’t melt what it can’t see. Check the bevel angle, land, and gap against the drawing before you tack, and tack so the gap doesn’t close as the joint shrinks. If you cannot slip the root pass into the gap cleanly, open it.

Clean to Bare Metal — Both Sidewalls

Mill scale, rust, oil, paint, and moisture all sit between your puddle and the base metal. Grind both bevel faces and at least an inch of adjacent plate down to bright metal, then wipe with a clean solvent and let it flash off. On aluminium, remove the oxide layer immediately before welding; it reforms fast and is a fusion killer.

Preheat Thick Sections, Watch Interpass

Thick plate acts like a heat sink, pulling heat from the puddle faster than the arc can put it in. Preheat heavy sections and high-carbon or alloy steels to the temperature your procedure specifies, and hold interpass temperature in range — too cold means no fusion, too hot means poor toughness.

Pick a Joint the Gun Can Actually Reach

Geometry has to suit the process. A fat stick electrode or a bulky MIG nozzle needs clearance to angle into the corner; if the bevel is too narrow or the fit-up too deep, you simply cannot point the arc at both sidewalls. Test the approach before tacking, and adjust the prep rather than forcing a bad angle.

Travel Speed: Let the Puddle Wet the Sidewalls

Running too fast leaves the puddle chasing the arc instead of wetting into the joint. Slow down until you see the molten pool bite into both sidewalls and wash out smoothly at the toes. The trick is consistency — dragging slowly through one section and racing through the next is how fusion defects get buried.

Weave and Dwell: More Is Not Less

Wide, slow weaves feel productive, but they let the arc linger in the middle and under-heat the edges. Excessive dwell and over-wide weaving are classic cold lap generators, which is why qualified procedures cap weave width and travel. Stay inside those limits; if you need more coverage, add passes, not oscillation.

Work Angle and Arc Length

Point the electrode into the corner so heat splits between both plates, and keep the arc short. A long arc loses focus and heat density, and a lazy angle aims at one sidewall while the other stays cold. Watch the puddle shape — it should be symmetrical and wetting both edges.

Keep the Equipment Honest

Worn contact tips, a kinked liner, a loose or dirty ground clamp, or undersized cables all rob current and destabilize the arc — the result is a colder weld than the machine displays. Set parameters with a meter, not your ear, and check them at the stinger. The full welding practice guide covers this discipline well.

Control the Environment

Drafts and wind strip shielding gas from the arc, degrading stability and inviting fusion and porosity defects. Build screens, close doors, and keep damp electrodes and wire out of the joint — moisture flashes to steam and blows the puddle around. On flux-cored work, slag can mask trouble, so fit-up and travel discipline matter even more.

Note: A convex, rope-like bead with a distinct toe line is a warning sign, not a style choice. It usually means the parameters are too cold for the material, and poor fusion at the toes often travels with it. Flatten the bead by adding heat and slowing down, or cut it out.

None of this is glamorous, and none of it happens at the puddle. It happens at the bench, with a grinder and a gap gauge. Get the prep right, and incomplete fusion stops being a gamble.

Common Mistakes Welders Make (and How to Avoid Them)

Almost every incomplete fusion failure traces back to a handful of repeatable habits. Here are the ones I see most, framed as the error, why it fails, the fix, and the visual cue that gives it away.

Sprinting with travel speed. Cranking speed to “go faster” pulls the arc ahead of the puddle, so sidewalls never melt. Match travel to the puddle. Clue: wagon tracks and a ropey bead resting on top of the plate.
Cutting amperage to stop burn-through. Thin metal doesn’t need a colder arc — it needs faster travel or a backing bar. Lowering heat without changing technique parks unfused metal on the surface. Clue: high crown, sharp unfused toes.
Skipping bevel or cleaning because “the arc will burn through scale.” It won’t. Mill scale, rust, oil, and paint keep base metal from reaching melting temperature at the interface. Grind to bright metal first. Clue: intermittent lack of fusion along the joint face.
Long arc length or stickout. Both drop voltage and rob penetration. Keep a tight arc, and check that worn or wrong contact tips aren’t letting stickout wander. Clue: spattery, wide, shallow bead.
Wrong polarity. Self-shielded flux-cored runs electrode-negative (straight polarity); switch to solid wire or gas-shielded FCAW and polarity flips. Backwards polarity causes fusion and appearance problems — see DCEN vs DCEP. Clue: erratic arc, poor sidewall wetting.
Skipping interpass slag removal. Slag along a bead edge makes wagon track indications, often beside unfused sidewalls on multi-pass FCAW and SAW work. Chip and brush every pass. Clue: dark lines tracing the previous toe.
Trusting a pretty cap. Ripples prove nothing underneath. The “cold and pretty” weld — smooth and never bonded — is the most dangerous one in the shop. Compare against our guide to good and bad MIG welds.
Welding over moisture, oil, or paint. Contaminants burn off and block fusion. Clean and dry the joint first. Clue: porosity and popping paired with unbonded edges.
No preheat on thick or restrained joints. Heavy sections wick heat away faster than the arc adds it. Preheat per the procedure. Clue: sluggish puddle and lack of fusion on first passes.
Over-weaving to fill a wide gap. Long side-to-side motion cools the puddle and lays metal on unmelted walls. Fix the fit-up instead. Clue: undercut beside an unfused area.
Wrong electrode for the position or joint. A flat-position rod dragged vertical-up, or a mismatch to joint geometry, fights gravity and fusion. Match classification and position to the job.
Ignoring the WPS because “I’ve always done it this way.” Following the qualified WPS is a code requirement in most structural, pressure, and pipeline work under standards like AWS D1.1. It exists to prevent defects like this one.
Assuming one NDT method sees everything. No single technique catches every orientation of a planar defect. Layer methods and follow the procedure.

Two traps deserve their own warning. The “cold and pretty” weld gets read as a pass, so nobody looks. And “just fill it with more passes” doesn’t work: depositing extra beads over an unbonded bead does not repair lack of fusion — the unbonded interface stays until it’s ground out and rewelded. You can stack five flawless-looking passes on a seam that was never joined.

Short Note: Photograph suspect welds before repair. A dated photo of toe lines, wagon tracks, or the fit-up lets you trace the failure to process variables — polarity, travel speed, cleaning — instead of guessing after the evidence is ground away.

None of these mistakes require a bad welder — just a rushed moment. The fix is usually slower travel, cleaner metal, and a quick check against the procedure before the arc starts.

Repair, Prevention, and Building a Fusion-First Welding Culture

Everything in this guide points to one uncomfortable truth: incomplete fusion is a process outcome, not bad luck. The good news is that both repair and prevention are entirely within your control — and neither is complicated.

Repairing Lack of Fusion the Right Way

Proper repair requires removing the defect to sound metal — typically by grinding or arc gouging — before depositing new weld metal. Skipping that step and simply running a cap pass over the seam — the so-called “cap it and go” fix — only buries the unbonded plane deeper inside the joint where nothing can find it. Once the defect is excavated, confirm it with penetrant or magnetic particle testing, then reweld using a qualified procedure.

Most codes treat repairs as seriously as original welds. AWS D1.1 and ASME BPVC Section IX both require qualified welding procedures and qualified welders, and many construction codes go further, requiring requalification of the repair procedure itself in demanding applications. That is why repair records, NDT reports, and welder IDs should trace back to the joint.

Quick Note: A repair that isn’t inspected after welding is an assumption, not a fix. Re-inspect to the same acceptance criteria as the original joint.

Prevention Systems That Actually Work

Prevention starts before the arc ever strikes, and it is built from documented systems rather than good intentions:

WPS and PQR: A written Welding Procedure Specification backed by a Procedure Qualification Record defines what actually produces fusion.
Welder qualification: Performance qualification and re-certification intervals keep skills current. Pre-job mock-up practice on the real joint beats theory every time.
Equipment calibration: machines, meters, and fixtures drift. Verify them on schedule.
Consumables: inspect filler metals on receipt, and store low-hydrogen electrodes like E7018 in controlled ovens to prevent moisture pickup.
Environment: wind breaks, preheat, and shielding from drafts.
Positioning: jigs and positioners put the joint into a weldable position instead of forcing awkward angles.
Monitoring: watch heat input and travel speed in production.
NDT plan: match the method to the defect — ultrasonic testing where lack of fusion is the risk.

Pre-job training deserves special emphasis. Mock-up practice on the actual joint geometry, position, and process lets a welder dial in travel speed, work angle, and heat input before the part is at risk — a habit that pays off far more than correcting a rejected weld later.

Calibrated settings matter more than most people admit. Amperage, polarity, and wire-feed behavior all shape whether the arc wets into the sidewall — the same variables covered in our guides to E7018 amperage and flux-core polarity. Good weld defect troubleshooting always starts there.

Note: A fusion-first culture starts with the shop itself — layout, ventilation, and a positioner in the right place make good fusion easier. Our guide to planning welding facilities walks through the setup.

The Leadership Angle

Systems fail when culture does. Give welders stop-work authority without penalty, and run root-cause analysis instead of blame. Track defect types by process, joint, and welder so training and tooling fixes target the real problem. Review trends monthly — a spike in one joint type usually points to a fit-up or fixture issue, not a lazy welder.

Final Takeaway: Make Fusion a Habit

Incomplete fusion means the weld metal never bonded to the base metal or a previous pass. Its causes fall into four families: current and heat input, technique and travel angle, joint prep and fit-up, and consumables or shielding. Detection runs from visual inspection through PT, MT, UT, and destructive testing. Codes respond by demanding qualified procedures, qualified welders, and verified repairs.

So inspect your next cold-looking bead before you move on. Verify joint prep and heat input. Treat fusion as something you engineer, not something you hope for.

Grind or gouge to sound metal — never cap over a defect.
Confirm the excavation with PT or MT before rewelding.
Weld repairs to a qualified WPS using a qualified welder.
Store E7018 and other low-hydrogen electrodes in ovens.
Calibrate machines, meters, and fixtures on schedule.
Match the NDT method to the risk — UT for lack of fusion.
Track defects by joint and welder, then train to the trend.

Frequently Asked Questions

What is incomplete fusion in welding?

Incomplete fusion — often called lack of fusion — is a discontinuity where the weld metal fails to metallurgically bond to the base metal, the joint sidewall, or a previously deposited bead. Because the unbonded interface is sharp and planar, it behaves like a crack and can grow under load. It is generally a rejectable defect under most structural, pressure, and pipeline codes.

What is the difference between lack of fusion and lack of penetration?

Lack of penetration is a geometric shortfall: the weld does not extend through the full required root thickness. Lack of fusion is a bonding failure at an interface, such as the sidewall or between passes. They can occur together, but the causes and fixes are different — penetration problems point to joint design and root access, while fusion problems usually point to heat input, technique, and cleanliness.

Is incomplete fusion always visible from the outside?

No. Many incomplete fusion defects are internal and produce no surface indication, especially sidewall or interpass lack of fusion in multi-pass welds. Surface clues like a convex, cold-looking bead with a distinct toe line or ‘wagon tracks’ from trapped slag are warning signs, but the absence of visible defects does not prove the weld is fused.

Which NDT method detects incomplete fusion best?

Ultrasonic testing, particularly angle-beam and phased-array UT, is generally the most reliable method for the tight, planar geometry of lack of fusion. Radiography is much less sensitive to these defects because there is little density change across the unbonded interface. Magnetic particle testing helps for surface and near-surface defects in ferromagnetic material, and penetrant testing only works if the defect reaches the surface.

Why does my MIG weld look good but fail a bend test?

A pretty cap says nothing about sidewall or interpass fusion. A cold weld can produce smooth, uniform ripples while the weld metal merely sits against an unmelted sidewall. When the coupon is bent, that unbonded interface opens up and the test fails. The usual fixes are more heat input, a corrected gun angle, slower travel speed, and clean joint faces.

Can incomplete fusion be repaired by welding over it?

No. Adding passes over an unbonded interface does not create a bond — the defect remains under the new metal. Proper repair means removing the defect to sound metal, usually by grinding or arc gouging, verifying removal (often with magnetic particle or penetrant testing), and then rewelding to a qualified procedure.

What causes incomplete fusion at the root of a joint?

Root lack of fusion typically comes from a root gap or root face that is too tight or too thick, insufficient current, the wrong electrode angle, or travel speed that outruns the puddle. Contaminants and oxidation in the root also prevent wetting. Fixing fit-up and heat input is usually more effective than simply welding slower.

Which welding processes are most prone to incomplete fusion?

GMAW and FCAW are widely associated with cold lap and toe/sidewall lack of fusion, especially in short-circuit transfer or with poor gun angles. SMAW suffers when current is too low and travel too fast, TIG when heat is inadequate or the sidewalls are not wetted, and SAW when bead alignment, flux handling, or current settings are wrong. Any process can produce it if heat input and joint access are inadequate.

How do codes treat incomplete fusion?

Most structural, pressure, and pipeline codes treat lack of fusion as a rejectable discontinuity, though the exact acceptance criteria depend on the governing specification and the service. AWS D1.1 covers structural steel, ASME BPVC Section IX governs qualification while ASME B31.3 and Section VIII set fabrication acceptance, and API 1104 governs pipeline welding. Always check the code that applies to your job.

What is the quickest way to reduce incomplete fusion in production?

The fastest wins are usually better joint prep and fit-up, verified heat input, and disciplined travel speed and gun angle. Confirm your polarity and wire or electrode selection, clean the joint to bare metal, control preheat and interpass temperature, and make sure machines and ground connections are delivering the current your procedure calls for.